phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-06-28 02:53:30 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat: Structured output for Responses API (#2324)

Browse source 
# What does this PR do?

This adds the missing `text` parameter to the Responses API that is how
users control structured outputs. All we do with that parameter is map
it to the corresponding chat completion response_format.

## Test Plan

The new unit tests exercise the various permutations allowed for this
property, while a couple of new verification tests actually use it for
real to verify the model outputs are following the format as expected.

Unit tests:

`python -m pytest -s -v
tests/unit/providers/agents/meta_reference/test_openai_responses.py`

Verification tests:

```
llama stack run llama_stack/templates/together/run.yaml
pytest -s -vv 'tests/verifications/openai_api/test_responses.py' \
  --base-url=http://localhost:8321/v1/openai/v1 \
  --model meta-llama/Llama-4-Scout-17B-16E-Instruct
```

Note that the verification tests can only be run with a real Llama Stack
server (as opposed to using the library client via
`--provider=stack:together`) because the Llama Stack python client is
not yet updated to accept this text field.

Signed-off-by: Ben Browning <bbrownin@redhat.com>
This commit is contained in:
Ben Browning 2025-06-03 17:43:00 -04:00 • committed by GitHub
parent c70ca8344f
commit 8bee2954be
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
8 changed files with 323 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

36
tests/verifications/openai_api/test_responses.py
View file

@ -546,3 +546,39 @@ async def test_response_streaming_multi_turn_tool_execution(
assert expected_output.lower() in final_response.output_text.lower(), (
f"Expected '{expected_output}' to appear in response: {final_response.output_text}"
)
@pytest.mark.parametrize(
"text_format",
# Not testing json_object because most providers don't actually support it.
[
{"type": "text"},
{
"type": "json_schema",
"name": "capitals",
"description": "A schema for the capital of each country",
"schema": {"type": "object", "properties": {"capital": {"type": "string"}}},
"strict": True,
},
],
)
def test_response_text_format(request, openai_client, model, provider, verification_config, text_format):
if isinstance(openai_client, LlamaStackAsLibraryClient):
pytest.skip("Responses API text format is not yet supported in library client.")
test_name_base = get_base_test_name(request)
if should_skip_test(verification_config, provider, model, test_name_base):
pytest.skip(f"Skipping {test_name_base} for model {model} on provider {provider} based on config.")
stream = False
response = openai_client.responses.create(
model=model,
input="What is the capital of France?",
stream=stream,
text={"format": text_format},
)
# by_alias=True is needed because otherwise Pydantic renames our "schema" field
assert response.text.format.model_dump(exclude_none=True, by_alias=True) == text_format
assert "paris" in response.output_text.lower()
if text_format["type"] == "json_schema":
assert "paris" in json.loads(response.output_text)["capital"].lower()